出于必要，这家德国消费品公司将生成式人工智能作为其五年数字化转型使命的核心。
四年前，德国跨国公司汉高发现自己处于十字路口。与消费品（CPG）行业的许多老牌企业一样，汉高在接受数字技术方面进展缓慢，导致这家拥有147年历史的公司与客户不断变化的需求之间的脱节越来越大。
正如汉高首席数据官迈克尔·尼尔斯（Michael Nilles）所说，到2019年，马克·安德森（Marc Andreessen）关于“软件正在吞噬世界”的声明在CPG行业已经成真，汉高面临落后的风险。
“我们认真对待它，并说我们需要软件、数据和人工智能功能，”当时担任 CDIO 角色的 Nilles 说。“我们需要加倍努力，大规模提高内部组织的技能，并从外部获取人才，我们称之为注入人才。”
为了实现其愿景，汉高制定了一个五年战略路线图，其中包括重组IT组织，创建一个新的数字部门，将CIO和CDO的风险投资活动整合到一个屋檐下，并在柏林、上海、班加罗尔和美国等中心建立全球创新中心。
如今，这些努力正在取得成果，使汉高成为采用生成式人工智能的领先公司之一，不仅可以优化其业务，还可以将其作为其未来战略愿景的核心组成部分。
一、为变革做好准备
汉高总部位于杜塞尔多夫，拥有50,000多名员工，不仅仅是一家CPG企业。其工业B2B部门专注于乐泰等粘合剂技术，而其B2C消费品部门则拥有Dial和Purex等品牌。
但为了实现汉高的数字愿景，Nilles需要吸引数据科学家、数据工程师和人工智能专家加入一个他们可能不会关注的行业。正如尼尔斯所说，人才是一场“核战争”，但对汉高有利的一个因素是，公司有有意义的问题需要解决。
“我认为我们很幸运，因为我们有有趣的行业问题需要解决，”Nilles说。“为大型科技公司工作的基础人工智能超级大师可能对加入像汉高这样的公司不感兴趣，但他也可能不是合适的人选。我们正在做的是找到那些喜欢用技术解决重大行业问题的人。
Nilles对汉高计划的另一个关键组成部分是建立强大的战略合作伙伴关系。汉高已经与SAP建立了合作关系，Nilles选择通过全力支持SAP的业务技术平台（BTP）并与软件公司密切合作进行共同创新来深化这种关系。该合作伙伴关系被称为 Digital Leapfrog，其首批成果之一是人工智能驱动的贸易促进管理 （TPM） 和贸易促进优化 （TPO） 工具。
TPM 和 TPO 是 CPG 领域的关键学科，涉及管理和优化与零售商进行的所有促销活动，从折扣到扣除和付款。像汉高这样的大型快消品公司可以在贸易推广上花费数十亿美元，因此要做到这一点有很多利害关系。
Nilles 说：“多年来，大多数行业都没有这样做，因为该领域没有真正的标准软件可用。“这是一个有点棘手的计算机科学问题。你有一个高度复杂的数据模型，你需要大量的计算，你需要有一个真正智能的UI。
二、通过必要性进行发明
由于在市场上找不到解决方案，汉高决定建立一个解决方案。汉高和SAP联合创新团队密切合作，构建和扩展了该工具，该工具必须能够处理超过20亿个计划节点。当时，该团队专注于传统人工智能，使用机器学习功能构建推荐引擎，帮助最终用户即时执行 TPO。
“在贸易促进管理中，边际优化对业务有巨大的影响，包括收入和利润，”Nilles说。
然而，与此同时，SAP正在为SAP Analytics Cloud开发一项新功能：Just Ask，它将Gen AI应用于搜索驱动的分析。该团队尝试将 Just Ask 与 TPO 工具一起使用，并很快将其视为实现该工具承诺的关键。
Nilles 说：“贸易促销的整个使用仍然是一个复杂的工具，对于大客户经理来说，它已经变得更加直观。“大客户经理或销售人员正在查看贸易促销数据，它给出了非常好的提示。这在传统的机器学习中是不可能的。有了 AI 世代，AI 功能已经被非数据科学博士的人更广泛地使用。
Nilles 指出，过去，与零售商围绕产品开展活动可能需要几个月的时间来敲定细节，例如要提供的适当折扣。现在，客户经理可以走进会议，使用带有自然语言的工具或多或少地实时探索选项，并准备好第二天分享的完成计划。到下周，完整的活动可以准备好在全国的商店中销售。
三、增长的催化剂
该工具已经取得了巨大的成功，但Nilles认为，Gen AI可以在汉高发挥更大的战略作用。该公司一直在努力将其CPG研发的所有数据转换为大型语言模型（LLM），Nilles表示，这将成为汉高开发新产品的巨大加速器，并帮助汉高在其领域发挥领导作用，因为人工智能颠覆了整个行业。
“我们相信会有新的垂直LLM出现，甚至可能是特定领域的微型垂直LLM，”他说。“我们相信，为正确的事情拥有LLM将是一个巨大的竞争优势，如果我们不这样做，我们将受到其他人的严重威胁和危害。
例如，想象一下网球设备市场。今天，有兴趣购买网球器材的人可能会去专门从事这项运动的网站。但随着人工智能的发展，消费者可能会去Meta或腾讯等公司，并在提示中提出查询。例如，无论谁构建了微信用来回答该问题的LLM，都将在市场上占据优势。
Nilles说：“如果我们是第一个，我们就设定了市场，当整个价值链重新布线时，我们有权在谈判桌上发挥作用。
翻译：
Henkel embraces gen AI as enabler and strategic disruptor
Out of necessity, the German consumer packaged goods company is putting generative AI at the core of its five-year digital transformation mission.
Four years ago, German multinational Henkel found itself at a crossroads. Like many incumbents in the consumer packaged goods (CPG) industry, Henkel was slow to embrace digital technologies, resulting in a widening disconnect between the 147-year-old company and the changing needs of its customers.
As Henkel CDIO Michael Nilles puts it, by 2019, Marc Andreessen’s pronouncement that “software is eating the world” had come true for the CPG sector, and Henkel was at risk of falling behind.
“We took it seriously and said we need to have software, data, and AI capabilities,” says Nilles, who signed on to the CDIO role at the time. “We needed to double down on the massive upskilling of the internal organization and also acquire — we call it injecting — talent from the outside.”
To achieve its vision, Henkel laid down a five-year strategic roadmap that involved reshuffling the IT organization, creating a new digital unit, consolidating CIO and CDO venture activities under one roof, and building global innovation centers in hubs like Berlin, Shanghai, Bangalore, and the US.
Today, those efforts are coming to fruition, positioning Henkel among the leading wave of companies adopting generative AI to not only optimize its businesses, but use it as a core building block of its strategic vision for the future.
1. Gearing up for change
Headquartered in Dusseldorf with 50,000-plus employees, Henkel is more than a CPG enterprise. Its industrial B2B arm focuses on adhesives technologies, like Loctite, while its B2C consumer goods arm owns brands such as Dial and Purex.
But to achieve Henkel’s digital vision, Nilles would need to attract data scientists, data engineers, and AI experts to an industry they might not otherwise have their eye on. As Nilles says, there’s a “nuclear war” for talent, but one factor in Henkel’s favor is that the company has meaningful problems to solve.
“We’ve been lucky, I think, because we have interesting industry problems to crack,” Nilles says. “The foundational AI super guru who’s working for big tech is maybe not interested in joining a company like Henkel, but he’s also probably not the right guy for that. What we’re doing is finding the guys who like to crack big industry problems with technology.”
Another key component of Nilles’ plan for Henkel has been to build strong strategic partnerships. Henkel already had a relationship with SAP, and Nilles opted to deepen that relationship by going all-in on SAP’s Business Technology Platform (BTP) and working closely with the software company on co-innovation. The partnership was dubbed Digital Leapfrog, and one of its first fruits was an AI-powered trade promotion management (TPM) and trade promotion optimization (TPO) tool.
TPM and TPO are key disciplines in the CPG space that involve managing and optimizing all promotional activities conducted with retailers, from discounts to deductions and payments. Big CPG companies like Henkel can spend billions on trade promotion, so there’s a lot at stake in getting it right.
“Most of the industry wasn’t doing that right for many years because there wasn’t really a standard software available in the space,” Nilles says. “It’s a bit of a tough computer science problem. You have a highly complex data model, you need a lot of compute, and you need to have a really smart UI.”
2. Invention through necessity
Unable to find a solution in the marketplace, Henkel decided to build one. The Henkel and SAP co-innovation team worked closely together to build and scale the tool, which had to be able to handle more than two billion planning nodes. At the time, the team was focusing on traditional AI, using machine learning capabilities to build a recommendation engine that could help end users perform TPO on the fly.
“In trade promotion management, marginal optimization has a huge impact on the business, both top and bottom line,” says Nilles.
However, at the same time, SAP was working on a new feature for the SAP Analytics Cloud: Just Ask, which applies gen AI to search-driven analytics. The team experimented with using Just Ask with the TPO tool, and quickly saw it as the key to fulfilling the tool’s promise.
“The whole use of trade promotion, which is still a complicated animal, has become much more intuitive for the key account managers,” Nilles says. “The key account manager or the salesperson is looking at the trade promotion data and it’s giving really great hints. This just wasn’t possible with traditional machine learning. With gen AI, the AI capabilities have become much more widely usable by people who aren’t PhDs in data science.”
In the past, Nilles notes that creating a campaign with retailers around a product could take months to hammer out details, like the proper discounts to offer. Now account managers can walk into a meeting, use the tool with natural language to explore options in more or less real-time, and have a finished plan ready to share the next day. By the next week, the complete campaign can be ready to go in stores nationwide.
3. A catalyst for growth
The tool has already been a big success, but Nilles believes gen AI can play an even bigger strategic role at Henkel. The company has been hard at work translating all the data from its CPG R&D into a large language model (LLM), which Nilles says will be a huge accelerator for Henkel to develop new products and help position Henkel to take a leadership role in its space as gen AI disrupts the industry.
“We believe there’ll be new vertical LLMs emerging, and maybe even micro-vertical ones that are domain-specific,” he says. “We believe having LLMs for the right things will be a huge competitive advantage, and that if we don’t do it, we’ll be seriously threatened and jeopardized by others.”
Imagine, for instance, the tennis equipment marketplace. Today, someone interested in buying tennis equipment might go to a website that specializes in the sport. But with the growth of gen AI, consumers might go to the likes of Meta or Tencent, and put a query in a prompt. Whoever built the LLM that WeChat, for example, uses to answer that query is going to have an advantage in the marketplace.
“If we’re one of the first, we set the market and we have the right to play at the table when there’s a rewiring of the whole value chain,” Nilles says.
本文由数字化转型网（www.szhzxw.cn）转载而成，来源于CIO.COM；编辑/翻译：数字化转型网宁檬树。
免责声明: 本网站(http://www.szhzxw.cn/)内容主要来自原创、合作媒体供稿和第三方投稿，凡在本网站出现的信息，均仅供参考。本网站将尽力确保所提供信息的准确性及可靠性，但不保证有关资料的准确性及可靠性，读者在使用前请进一步核实，并对任何自主决定的行为负责。本网站对有关资料所引致的错误、不确或遗漏，概不负任何法律责任。
本网站刊载的所有内容(包括但不仅限文字、图片、LOGO、音频、视频、软件、程序等) 版权归原作者所有。任何单位或个人认为本网站中的内容可能涉嫌侵犯其知识产权或存在不实内容时，请及时通知本站，予以删除。
